Are patterns of pandemic spread, its determinants, and effects of public health interventions are similar across pandemics?
Estimate excess mortality for pandemics in 1890, 1918 and 2020 per district, age groups and sex
Comparing of spatial pattern between the pandemics
Investigate the determinants of spread in the context of different co-factors ( Urbanization, GIP per capita etc.)
Data
Collected and digitalized from Kaspar Staub’s team
Russian flu: 1879 - 1895
Spanish fl: 1908 - 1925
Covid19 : 2020
Population
Modern data population for all districts and all age groups and sex available
Census 1888 and 1910 population data for all districts and age groups and sex, but 1900 and 1920 are not collect -> might a problem
Census 1880, 1888, 1900, 1910, 1920 census data for all districts
Estimation:
Interpolation for total and each districts (census 1880, 1888, 1900, 1910, 1920)
Calculation of age distribution for 1888 and 1910
Take age distribution of 1888 to interpolate population between 1880 to 1900 and age distribution from 1910 to interpolate population between 1900 and 1920
Maybe a student from Kaspar will also collect detailed census data from 1900 and 1920, would be a bit more precise then
Maps
All districts are harmonised so that 1890, 1918 and 2020 have the same districts.
Many districts from 1879 - 1920 have been combined so that they are the same as 2014-2020.
Schauffhausen is only one district, as death data is only available for Schaffhausen as a whole.
In Solothurn, the districts are merged as in 1876 - 1920.
New shapefiles created via QGIS to have one map for all years (to make them comparable).
Methods
Prior: Default Gamma distribution
Estimation of the expected death counts for each district, sex and age
Poisson Regression
Natural logarithm of the population in each district (sex, age) was used as the model’s offset term
Estimation based on the mortality trend of the previous 4 years (1890 only 4 years possible)
Pandemic years 1890, 1918 and 2020 are excluded to estimate expected mortality (high mortalities in these years, what would be observed without a pandemic?)
Bootstrapping to address the uncertainty in observed death counts and to provide a prediction interval (PI) for the predicted mortality(resampled N = 1000)
Excess mortality = observed death counts – expected death counts
Excess mortality is shown relatively in percentage (Excess mortality/expected death counts)
To compare the districts of each pandemic year the relative excess mortality was normalized
To find pattern and clusters: Moran’s I statistics ( LISA: Local Indicators of Spatial Association) were used for mapping statistically significant local clusters
Preliminary Results
Total
Relative yearly numbers of excess deaths
Maps
LISA (Local Indicators of Spatial Association)
Sex
Maps
LISA (Local Indicators of Spatial Association)
Age
Only two age groups 0-69 and >=70.
More precise age groups would lead to a lot of zeros in small districts
Even with two age groups, zeros in some districts
Maps
LISA (Local Indicators of Spatial Association)
Next steps, Points to be discussed
Bayesian approaches in hierarchical modelling (INLA)
Lot of zeros, especially for age groups
Bayesian approaches in hierarchical modelling (INLA) to investigate the spatial pattern of excess mortality per district
Groups
Too many age groups lead to too many zeros -> too many zeros are also a problem with INLA
Maybe only 0-69 and >70 or 0-40, 41-69, >70
Further Co-factors (I have to discuss with Kaspar):
Urbanization
Infant mortality rates as a proxy for health index
Public health intervention for each district (canton)
GDP per capita as proxy for SES
Population density (population/km2)
Proportion of children, 5–15 y (as school-age children are thought to drive influenza transmission)